The EMBO Journal Vol.17 No.8 pp.2353–2358, 1998
Structure of wild-type yeast RNA polymerase II and
location of Rpb4 and Rpb7
Grant J.Jensen, Gavin Meredith,
David A.Bushnell and Roger D.Kornberg1
Department of Structural Biology, Stanford University School of
Medicine, Stanford, CA 94305, USA
1Corresponding
author
The three-dimensional structure of wild-type yeast
RNA polymerase II has been determined at a nominal
resolution of 24 Å. A difference map between this
structure and that of the polymerase lacking subunits
Rpb4 and Rpb7 showed these two subunits forming
part of the floor of the DNA-binding (active center)
cleft, and revealed a slight inward movement of the
protein domain surrounding the cleft. Surface plasmon
resonance measurements showed that Rpb4 and Rpb7
stabilize a minimal pre-initiation complex containing
promoter DNA, TATA box-binding protein (TBP),
transcription factor TFIIB and the polymerase. These
findings suggest that Rpb4 and Rpb7 play a role in
coupling the entry of DNA into the active center cleft
to closure of the cleft. Such a role can explain why
these subunits are necessary for promoter-specific transcription in vitro and for a normal stress response
in vivo.
Keywords: electron crystallography/Rpb4/Rpb7/
transcription
Introduction
Nucleic acid polymerase structure has been studied by
both X-ray and electron crystallography. To date, only the
smaller, single subunit polymerases have been subjected
to X-ray analysis, including the bacteriophage T7 RNA
polymerase, which is the only RNA polymerase whose
structure is known to atomic resolution (Sousa et al.,
1993). Lower resolution structures of several multisubunit
polymerases have been determined by electron crystallography, including core and holo- Escherichia coli RNA
polymerase (Darst et al., 1989; Polyakov et al., 1995),
yeast RNA polymerase I (Schultz et al., 1990) and a
mutant form of yeast RNA polymerase II (Darst et al.,
1991a). All polymerase structures obtained by both X-ray
and electron crystallography show a large cleft appropriate
in size for binding duplex DNA, and further appear to
contain a mobile arm allowing open and closed conformations of the cleft, presumably permitting entry and retention
of DNA. In the specific case of RNA polymerase II from
the yeast Saccharomyces cerevisiae, which is the subject
of this work, it has been shown that nucleic acid does in
fact occupy this cleft in a paused elongation complex
(C.L.Poglitsch, G.Meredith, A.Gnatt, G.J.Jensen, W.-h.
Chang and R.D.Kornberg, in preparation), and that the arm
domain enclosing the cleft is mobile (Asturias et al., 1997).
RNA polymerase II, which catalyzes the transcription
of all protein-encoding genes, is composed in yeast of 12
unique subunits denoted Rpb1–Rpb12. Two of these
subunits, Rpb4 and Rpb7, form a dissociable subcomplex
(Dezelee et al., 1976; Ruet et al., 1980; Edwards et al.,
1991), present in substoichiometric amounts during exponential growth (Choder and Young, 1993). This complex
has been implicated in the stress response and in the
initiation of transcription (Woychik and Young, 1989;
Edwards et al., 1991; Choder and Young, 1993). Human
homologs of Rpb4 and Rpb7 have been identified and
have also been shown to form a dissociable subcomplex
of human RNA polymerase II (Khazak et al., 1998).
Heterogeneity of RNA polymerase II preparations
resulting from substoichiometric amounts of Rpb4 and
Rpb7 interfered with crystallization and was circumvented
by the use of a mutant strain in which the RPB4 gene
was deleted (Darst et al., 1991b). Enzyme isolated from
this strain lacks both Rpb4 and Rpb7 (and is therefore
denoted Δ4/7) and forms large two-dimensional crystals
on lipid layers, enabling structure determination at low
resolution by electron crystallography (Darst et al., 1991a).
More recently, it has been shown that polymerase purified
from cells in stationary phase has a full complement of
Rpb4 and Rpb7 (Choder and Young, 1993), permitting
the isolation of homogeneous, 12 subunit enzyme, which
we refer to as wild-type. We report here on the crystallization and structure determination of this wild-type polymerase, and on the localization of Rpb4 and Rpb7 by
direct difference analysis between the wild-type and the
Δ4/7 polymerases. Finally, we discuss the functional
insight provided by the location of Rpb4 and Rpb7 in
light of their roles in the stress response and in the
initiation of transcription.
Results
A three-dimensional reconstruction of wild-type RNA
polymerase II was calculated from the computed diffraction patterns of two-dimensional crystals grown on lipid
layers and imaged in the electron microscope. Data collection for the wild-type polymerase reconstruction was
improved by using a wire loop to transfer crystals to the
specimen grid and by washing the grids with dilute
detergent solution to remove excess lipid (Asturias and
Kornberg, 1995). To ensure an appropriate comparison,
data for the Δ4/7 enzyme were recollected using these
same procedures, and an improved reconstruction was
calculated. Wild-type and Δ4/7 polymerase data sets were
derived from untilted images and 64 or 69 tilted images,
respectively, recorded at various tilt angles of up to 59°
to the incident electron beam. Crystals formed in plane
2353
G.J.Jensen et al.
Fig. 1. Reciprocal space coverage in the wild-type and Δ4/7 reconstructions. For both the wild-type and Δ4/7 reconstructions (left and right sides of
the figure, respectively), four statistics are presented in each of 16 regions of reciprocal space. The 16 regions are delimited by resolution and tiltangular boundaries. In each region, the number in the upper left corner is the percentage of possible structure factors that were measured adequately
enough to be sampled. Because, in this technique, measurements are dispersed non-uniformly along lattice lines, whether or not a structure factor is
measured adequately to justify sampling can be somewhat subjective. Thus we report in the upper right corner a statistic that we call the redundancy
of measurement, which is defined as the average number of measurements recorded that lie closer than half the sampling interval (⫾ 0.0016 Å–1 in
this case) to each sampled structure factor. Note that only measurements in sampled regions of space are included. The number in the bottom left
corner is the average phase error, defined as the average phase difference between measurements within 1/(4T) in z*, where T is an estimate of the
thickness of the molecule in the direction perpendicular to the plane of the crystal, and z* is that direction in reciprocal space (here, T ⫽ 100 Å).
Again, only measurements in sampled regions of space are included. The number in the bottom right corner is the number of phase comparisons
used to determine the average phase error. We chose to report differences between actual phase measurements rather than differences between
measurements and a curve-fit lattice line because in regions of space with few measurements, the lattice lines are relatively free to match the
measurements closely.
group c2, and both data sets allowed sampling of ⬎50%
of all possible reflections to 24 Å resolution, with a
missing cone of 45°. The average phase errors of individual
measurements along lattice lines within 0.0025 in z* (the
direction perpendicular to the crystal plane) were 19° and
15° for the wild-type and the Δ4/7 structures, respectively
(Figure 1).
Both the wild-type and the new Δ4/7 polymerase
structures (Figure 2) were remarkably similar to that of
the Δ4/7 polymerase reported previously (Darst et al.,
1991a). For the purpose of discussion, the surface shown
by the right member of the pair of molecules is referred
to as the ‘front’ side, and the surface of the left member
is termed the ‘back’ side. Among many features common
to the two structures, the 25 Å cleft (labeled ‘C’ in Figure
2A), which passes from the front to the back side, contains
nucleic acid in a halted elongation complex (C.L.Poglitsch,
G.Meredith, A.Gnatt, G.J.Jensen, W.-h.Chang and R.D.
Kornberg,, in preparation) and is believed, by analogy to
single subunit polymerases, to harbor the active center
and bind DNA. Another feature seen in both forms of
polymerase, but not discussed previously, was a groove
on the front face that extended from the DNA-binding
cleft towards the upper left corner (molecule on the right
in Figure 2A). In the Δ4/7 polymerase, this groove
appeared enclosed at some points, forming a perforated
tunnel. The groove was appropriate in size for binding
2354
single-stranded nucleic acid, and was met at the upper
left corner of the molecule on the right in Figure 2A by
the end of another groove lying on the back face, similar
in size, that also originated in the nucleic acid-binding
cleft, forming a nearly continuous ring around the upper
region of the polymerase.
In contrast to the previously published RNA polymerase
II structure, in which the arm of protein density enclosing
the cleft (labeled ‘A’ in Figure 2A) extended downwards,
the arm appeared as a discrete domain in both enzymes
at the contour level used here, and was connected with
the bottom half of the molecule at lower contour levels.
A further difference was in the absence here of a tunnel
35 Å in length extending through the previous structure.
These differences can be accounted for by improved data
quality, due to the transfer and washing techniques used
and to the nearly 2-fold increase in the amount of data
collected.
To reveal differences between the wild-type and Δ4/7
structures, the data sets were trimmed to analogous limits,
scaled, aligned and subtracted to produce a difference
map (wild-type minus Δ4/7). Because negative staining
reveals only a molecular envelope, difference ‘density’ in
areas devoid of stain within the hydrophobic core of a
protein or in areas filled with stain in the large intermolecular regions is unlikely to be meaningful. To display only
those differences that reasonably could be attributed to
Location of Rbp4 and Rbp7 in yeast RNA polymerase II
Fig. 2. (A) Three-dimensional reconstructions of the wild-type RNA polymerase II, the Δ4/7 RNA polymerase II and the difference between them.
Both forms of RNA polymerase crystallize as dimers related by a 2-fold rotation axis in the plane of the crystal. One dimer, composed of two
symmetry-averaged, identical monomers, is shown for each reconstruction. At the top, in blue, is the full reconstruction of the wild-type enzyme.
Below it, in yellow, appears the full reconstruction of the Δ4/7 enzyme. The bottom display presents the wild-type–Δ4/7 differences within the
boundaries described (see text) in red, contoured at 2.5 standard deviations above the mean, superimposed on a transparent view of the Δ4/7
reconstruction. The transparent view allows differences to be seen which are ‘buried’ in cavities or are inside the Δ4/7 reconstruction’s surface. Note
that on the right molecule, nearly all of the large, globular difference is covered by the front surface of the Δ4/7 reconstruction, while on the left
molecule (which shows the back side view), this same globular difference is directly visible. No negative differences are seen at 2.5 standard
deviations below the mean. Because negative stain is thought to partially penetrate a protein’s hydrophilic shell, each polymerase is contoured at the
level that encloses 70% of the expected volume. ‘C’ marks the 25 Å cleft that is occupied by nucleic acid in elongation complexes and is thought to
harbor the active center, while ‘A’ marks the mobile protein arm enclosing the cleft. The scale bar represents 50 Å. (B) Position of Rpb4 and Rpb7
with respect to the DNA-binding cleft. The top half of the figure is a slab section from the single Δ4/7 RNA polymerase II monomer on the bottom
right of (A), cut at the level of the cavity filled by Rpb4 and Rpb7, with differences at 2.5 standard deviations above the mean shown in red. The
bottom half of the figure was generated by superimposing slab sections of the wild-type (blue) and Δ4/7 (yellow) dimers. The slab extends from 14
Å away from the crystal’s midplane to 31 Å away. Because the slab is not centered on the midplane, the right and left halves are not mirror images,
having cut through the right and left monomers of (A) at different levels, and represent independent slab sections of the same structure. The right
section is a slab through the back end of the DNA-binding cleft where Rpb4 and Rpb7 are located, while the left section is a slab through the front
end of the DNA-binding cleft, which does not include Rpb4 and Rpb7. Both wild-type and Δ4/7 reconstructions are contoured here at a level that
encloses 100% of their expected volume to demonstrate that the cavity in the Δ4/7 enzyme corresponding to the location of Rpb4 and Rpb7 in the
wild-type enzyme is not generated artifactually by contouring at inappropriately high levels. In contrast to (A), the wild-type and Δ4/7
reconstructions in (B) have been trimmed, scaled and aligned for direct comparison (see Materials and methods).
protein subunits absent from the Δ4/7 polymerase and
present in the wild-type enzyme, the difference map was
set to zero at all points inside the contour which enclosed
50% of the expected volume of the Δ4/7 polymerase, and
at all points outside the contour which enclosed 150% of
the expected volume of the wild-type polymerase. When
contoured at 2.5 standard deviations above and below the
mean, this map showed only one major, positive difference
and three minor, positive differences (red densities in
Figure 2A).
The major difference was a globular density forming
part of the floor of the DNA-binding cleft, filling a cavity
2355
G.J.Jensen et al.
in the Δ4/7 structure not seen in the wild-type structure
(Figure 2A and B). The difference was centered ~10 Å
away from the midplane of the crystal towards the back
side. In the absence of any other large difference, we
identified this density with subunits Rpb4 and Rpb7, which
are known to form a single, dissociable subcomplex. The
volume of this globular density, as shown contoured at
2.5 standard deviations above the mean, was 31% of the
expected volume of Rpb4 and Rpb7. At slightly lower
contours, where the volume increased to 100% of that
expected for Rpb4 and Rpb7, the difference expanded in
all directions and connected with the smaller difference
directly below it.
The three minor differences between wild-type and
Δ4/7 structures may result from noise in the data or from
slight conformational changes induced by the presence of
Rpb4 and Rpb7. The small difference just inside the arm
domain (‘A’ in Figure 2A) was matched on the outside of
the domain by a more diffuse, negative difference at lower
contour levels (not shown). We suspect that these paired
positive and negative differences reflect a slight inward
movement of the arm domain caused by the presence of
Rpb4 and Rpb7. Elsewhere it has been shown that the
arm domain is mobile (Asturias et al., 1997), and the
movement seen here may represent a shift in the average
position of the domain rather than an abrupt transition
between two distinct conformations.
As a measure of significance, a difference map was
calculated in the two-sided plane group p1, which imposes
no symmetry, so that each structure factor derives from
only half as much data and each member of the pair of
polymerase molecules is determined independently. While
the difference map was more noisy than before, the largest
difference density by volume (contoured at 2.5 standard
deviations above the mean) in both polymerase molecules
was in the location assigned before to Rpb4 and Rpb7,
and these differences were the highest (most significant)
symmetric pair (data not shown). For further tests of
significance, difference maps were calculated in the twosided plane group c2 using alternative algorithms for
trimming, scaling and aligning the maps, leading in all
cases to the same conclusion about the location of Rpb4
and Rpb7 (data not shown). As a final control, we
performed the same difference analysis using the data
from the previous Δ4/7 structure, revealing a single large
difference at 2.5 standard deviations above the mean, with
the same location, general shape and approximate volume
as that shown here, as well as several minor differences,
including a peak just inside the arm domain (data not
shown).
Perhaps the best indication of the reliability of the data
is the close correspondence of the wild-type and Δ4/7
structures in regions remote from the Rpb4 and Rpb7
density (bottom of Figure 2B). Dissimilarities in the
outlines of the two molecules are due to Rpb4 and Rpb7
density, conformational changes caused by Rpb4 and
Rpb7, or noise. Thus the degree of dissimilarity in regions
away from Rpb4 and Rpb7 provides a high estimate of
the noise inherent in the technique. As the bottom of
Figure 2B shows two independent sections, both 17 Å
thick, about one-third of the total thickness of the molecules is presented in section here. Because the method of
electron crystallography prevents data collection within
2356
Fig. 3. Effect of Rpb4 and Rpb7 on the stability of the pre-initiation
complex. A mixture of 30 μg/ml of TBP, 38 μg/ml of TFIIB,
10 μg/ml of 33 bp double-stranded DNA containing the adenovirus
major late promoter sequence from –43 to –14 (5⬘-TTTCTGAAGGGGGGCTATAAAAGGGGGTGGGGG-3⬘) and 100 μg/ml of bovine
serum albumin was injected onto an 8000 resonance unit surface of
wild-type polymerase (top curve) or a 8500 resonance unit surface of
Δ4/7 polymerase (bottom curve). Interactions were measured with the
BIAcore Biosensor (BIAcore AB). The apparent association rate was
more than three times higher and the apparent dissociation rate five
times lower for the wild-type than for the Δ4/7 enzyme, indicating at
least a 15-fold greater affinity of the wild-type enzyme.
the ‘missing cone’ of reciprocal space, the resulting models
are most well determined in the midplane of the crystal,
and less well determined away from the midplane. Thus
the section of the molecule’s total thickness that is shown
here, which includes the density assigned to Rpb4 and
Rpb7, displays the parts of the molecular outlines that are
most appropriately used as internal controls.
Elsewhere, the presence of Rpb4 and Rpb7 was shown
to favor a closed conformation of the arm domain around
the DNA-binding cleft and, from the requirement for these
two subunits for the initiation of transcription (Edwards
et al., 1991), it was concluded that closure of the arm is
an essential step in the initiation process (Asturias et al.,
1997). The difference peak seen here suggesting a tighter
closure of the arm domain in the wild-type enzyme is
nicely consistent, and the location of the two subunits
between the two major domains that form the DNAbinding cleft suggests a mechanism whereby Rpb4 and
Rpb7 contact DNA in the cleft and couple its entry to
closure of the cleft.
We sought biochemical support for such a role of Rpb4
and Rpb7 by measurement of the affinities of wild-type
and Δ4/7 polymerases for DNA. The measurements were
made in the presence of TATA box-binding protein (TBP)
and TFIIB, as occurs in the natural initiation mechanism,
and which suffices for transcription in some circumstances
(Parvin and Sharp, 1993). Wild-type and Δ4/7 polymerases
were immobilized separately on sensor chips and tested
for interaction with an adenovirus major late promoter
DNA fragment–TBP–TFIIB complex by surface plasmon
resonance. The apparent association rate was more than
three times higher and the apparent dissociation rate five
times lower for the wild-type than for the Δ4/7 enzyme
(Figure 3), indicating at least a 15-fold greater affinity of
Location of Rbp4 and Rbp7 in yeast RNA polymerase II
the wild-type enzyme. We conclude that Rpb4 and Rpb7
stabilize a minimal pre-initiation complex.
Discussion
The discovery that yeast RNA polymerase II with a
full complement of Rpb4 and Rpb7 could be purified
from cells in stationary phase enabled us to form twodimensional crystals of wild-type polymerase suitable
for structural analysis to 24 Å resolution. As the crystals
were isomorphous to crystals formed with Δ4/7 enzyme,
direct difference analysis was performed, locating Rpb4
and Rpb7 in the floor of the DNA-binding cleft, between
the two major domains that form the cleft. A combination
of small positive and negative differences suggested
that Rpb4 and Rpb7 induce tighter closure of the arm
domain surrounding the cleft. The finding here and
elsewhere (Asturias et al., 1997) that Rpb4 and Rpb7
favor closure of the arm domain around the DNAbinding cleft is easily understood in light of the position
of Rpb4 and Rpb7 between the two major domains that
are brought together to form the cleft (Figure 2B).
The stabilization of polymerase–DNA complexes by
Rpb4 and Rpb7 can explain why these subunits are
important for the stress response and for the initiation
of transcription. While deletion of Rpb7 is lethal
(McKune et al., 1993), deletion of Rpb4 produces
strains that grow slowly and are sensitive to heat, cold
and starvation (Woychik and Young, 1989; Choder and
Young, 1993). Because the stress response depends on
the presence of transcriptionally engaged polymerase
molecules paused downstream of heat shock promoters
(Rasmussen and Lis, 1995), Rpb4 and Rpb7 may
facilitate the stress response by stabilizing paused
polymerases.
Similar reasoning can explain the role of Rpb4 and
Rpb7 in the initiation of transcription. While Δ4/7
RNA polymerase II is indistinguishable from wildtype polymerase in promoter-independent initiation and
elongation, it is inactive in promoter-directed initiation.
The activity of Δ4/7 polymerase in promoter-directed
initiation can be restored by addition of purified Rpb4
and Rpb7 or of activator proteins (Edwards et al.,
1991). We suggest that Rpb4 and Rpb7 sense entry of
DNA into the cleft and then induce and maintain cleft
closure, extending the lifetime of pre-initiation complexes
and allowing other factors to interact and perform their
roles. TFIIE may be a good example, as it has been
located bound to the arm around the cleft in the closed
conformation (Leuther et al., 1996). Activators may
obviate the need for Rpb4 and Rpb7 in the initiation
of transcription in vitro by either promoting the formation
of a larger number of relatively less stable initiation
complexes or by stabilizing the initiation complex as
do Rpb4 and Rpb7.
Materials and methods
Polymerase purification and crystallization
Wild-type and Δ4/7 polymerases were purified (Edwards et al., 1990)
and crystallized (Meredith et al., 1996) as described, with the use of
S.cerevisiae strains CB010 and CB010 ΔRpb4, respectively. Wildtype enzyme was derived from cells harvested during early stationary
phase. Polymerase composition was verified by SDS–polyacrylamide
gel electrophoresis followed by Coomassie staining and laser densitometry.
Crystal imaging and processing
Crystals were transferred to microscope grids, stained with uranyl
acetate and imaged with a Philips CM12 electron microscope operating
at 100 kV with a LaB6 filament at 35 000⫻ magnification using
low-dose techniques. Usually, three pictures were taken of each
crystal, including one untilted image. Suitable micrographs were
selected by optical diffractometry and scanned with either a PerkinElmer flatbed scanner or a Leaf Scanner with a 20 μm step size.
Images were processed in the two-sided plane group p1 or c2 (a ⫽
404 Å, b ⫽ 230 Å, and γ ⫽ 90° in both cases) as described (Darst
et al., 1991a) by standard methods (Amos et al., 1982; Henderson
et al., 1986). To estimate the resolution of each image, the radius at
which only half of the possible structure factors were measured with
signal-to-noise ratio ⬎2 was determined, and measurements more
than a few Ångstroms beyond this resolution were discarded as
unreliable. During the alignment and lattice line sampling, only
measurements with signal-to-noise ratios ⬎1.4 (IQ ⬍6 according to
Henderson et al., 1986) were used. Because the two monomers in
the p1 asymmetric unit were nearly identical, and because the average
phase error decreased from 20.1° to 19.2° upon c2 averaging for the
wild-type reconstruction, the figures were produced using symmetryaveraged, c2 data. Smooth curves were fit to the reciprocal space
lattice lines and were sampled at spacings corresponding to a unit
cell 300 Å thick to obtain three-dimensional Fourier terms. In regions
where there were no measurements within about half the sampling
interval along a lattice line or where the measurements were in poor
agreement, no structure factors were sampled. Reconstructions were
visualized with the XtalView software package (McCree, 1992), and
surface envelopes were generated by Synu (Hessler et al., 1992).
Predicted protein volumes were calculated using the estimate
1.21 Å3/Da.
Image processing was facilitated by three new FORTRAN programs
by G.Jensen, which are available upon request. The first finds initial
lattice vectors from an image transform for subsequent refinement,
and the second finds vertices for cutting out continuous regions of
crystal using the results of the ‘unbending’ process outlined in
Henderson et al. (1986). The third program collects the statistics
shown in Figure 1.
Difference analysis
Wild-type and Δ4/7 polymerase crystals were isomorphous, allowing
direct difference calculation after appropriate trimming, scaling and
aligning. Each set of structure factors contained ⬎50% of the expected
reflections to 24 Å resolution, with a missing cone of 45°. All
structure factors outside these limits were discarded. Amplitude scaling
was done in four separate bins (250–50 Å, 50–35 Å, 35–28 Å and
28–24 Å) to diminish any resolution-dependent differences in contrast
transfer by multiplying all the wild-type structure factor amplitudes
within a given bin by the appropriate scale factor to equalize the
mean amplitudes of the wild-type and the Δ4/7 sets of structure
factors. Reconstructions were aligned along the a and c axes of the
unit cell by defining the crystallographic 2-fold symmetry axis as the
b axis, while alignment along b was accomplished by minimizing
weighted phase differences between the sets of structure factors.
Differences were calculated directly in real space or equivalently in
reciprocal space by vector subtraction.
Surface plasmon resonance measurements
All proteins were diluted in running buffer (40 mM HEPES pH 7.6,
7.5 mM MgCl2, 120 mM potassium acetate, 0.005% Surfactant P20) from concentrated stock solutions prior to use. Wild-type and
Δ4/7 polymerases were immobilized in 100 mM sodium acetate pH
5.0 at 100 μg/ml on CM5 research-grade sensor chips with the use
of an amine-coupling kit (BIAcore AB). Blank surfaces were prepared
by introduction of buffer without protein. Non-covalently bound
proteins were removed by washing with running buffer containing
1 M potassium acetate. Mixtures of TBP, TFIIB, double-stranded
DNA containing adenovirus major late promoter sequence, and bovine
serum albumin were incubated on ice for 20 min and then injected
onto a polymerase surface at room temperature in running buffer at
a flow rate of 15 μl/min. Interactions were measured with the BIAcore
Biosensor (BIAcore AB). TBP and TFIIB were purified according to
Myers et al. (1997).
2357
G.J.Jensen et al.
Acknowledgements
We thank Steve Orlicky and Al Edwards for providing purified
recombinant Rpb4 and Rpb7 complex, and Seth Darst for helpful
advice. G.J.J. was supported by a Medical Scientist Training Program
grant (GM07365) provided by the National Institute of General
Medical Sciences at the NIH. This research was funded by NIH
grant AI21144 to R.D.K.
References
Amos,L.A., Henderson,R. and Unwin,P.N. (1982) Three-dimensional
structure determination by electron microscopy of two-dimensional
crystals. Prog. Biophys. Mol. Biol., 39, 183–231.
Asturias,F.J. and Kornberg,R.D. (1995) A novel method for transfer
of two-dimensional crystals from the air/water interface to specimen
grids. J. Struct. Biol., 114, 60–66.
Asturias,F.J., Meredith,G.D., Poglitsch,C.L. and Kornberg,R.D. (1997)
Two conformations of RNA polymerase II revealed by electron
crystallography. J. Mol. Biol., 272, 536–540.
Choder,M. and Young,R.A. (1993) A portion of RNA polymerase II
molecules has a component essential for stress response and stress
survival. Mol. Cell. Biol., 13, 6984–6991.
Darst,S.A., Kubalek,E.W. and Kornberg,R.D. (1989) Three-dimensional
structure of Escherichia coli RNA polymerase holoenzyme
determined by electron crystallography. Nature, 340, 730–732.
Darst,S.A., Edwards,A.M., Kubalek,E.W. and Kornberg,R.D. (1991a)
Three-dimensional structure of yeast RNA polymerase II at 16 Å
resolution. Cell, 66, 121–128.
Darst,S.A., Kubalek,E.W., Edwards,A.M. and Kornberg,R.D. (1991b)
Two-dimensional and epitaxial crystallization of a mutant form of
yeast RNA polymerase II. J. Mol. Biol., 221, 347–357.
Dezelee,S., Wyers,F., Sentenac,A. and Fromageot,P. (1976) Two forms
of RNA polymerase B in yeast. Eur. J. Biochem., 65, 543–552.
Edwards,A.M., Darst,S.A., Feaver,W.J., Thompson,N.E., Burgess,R.R.
and Kornberg,R.D. (1990) Purification and lipid-layer crystallization
of yeast RNA polymerase II. Proc. Natl Acad. Sci. USA, 87,
2122–2126.
Edwards,A.M., Kane,C.M., Young,R.A. and Kornberg,R.D. (1991)
Two dissociable subunits of yeast RNA polymerase II stimulate
the initiation of transcription at a promoter in vitro. J. Biol. Chem.,
266, 71–75.
Henderson,R., Baldwin,J.M., Downing,K.H., Lepault,J. and Zemlin,F.
(1986) Structure of purple membrane from Halobacterium halobium:
recording, measurement and evaluation of electron micrographs at
3.5 Å resolution. Ultramicroscopy, 19, 147–178.
Hessler,D. et al. (1992) Programs for visualization in three-dimensional
microscopy. Neuroimage, 1, 55–67.
Khazak,V., Estojak,J., Cho,H., Majors,J., Sonoda,G., Testa,J.R. and
Golemis,E.A. (1998) Analysis of the interaction of the novel RNA
polymerase II (pol II) subunit hsRPB4 with its partner hsPRB7
and pol II. Mol. Cell. Biol., 18, 1935–1945.
Leuther,K.K., Bushnell,D.A. and Kornberg,R.D. (1996) Twodimensional crystallography of TFIIB– and IIE–RNA polymerase
II complexes: implications for start site selection and initiation
complex formation. Cell, 85, 773–779.
McCree,D.E. (1992) A visual protein crystallographic software system
for XII/XView. J. Mol. Graphics, 10, 44–46.
McKune,K.,
Richards,K.L.,
Edwards,A.M.,
Young,R.A.
and
Woychik,N.A. (1993) Rpb7, one of two dissociable subunits of
yeast RNA polymerase II, is essential for cell viability. Yeast, 9,
295–299.
Meredith,G.D., Chang,W.H., Li,Y., Bushnell,D.A., Darst,S.A. and
Kornberg,R.D. (1996) The C-terminal domain revealed in the
structure of RNA polymerase II. J. Mol. Biol., 258, 413–419.
Myers,L.C., Leuther,K., Bushnell,D.A., Gustafsson,C.M. and
Kornberg,R.D. (1997) Yeast RNA polymerase II transcription
reconstituted with purified proteins. Methods, 12, 212–216.
Parvin,J.D. and Sharp,P.A. (1993) DNA topology and a minimal set
of basal factors for transcription by RNA polymerase II. Cell, 73,
533–540.
Polyakov,A., Severinova,E. and Darst,S.A. (1995) Three-dimensional
structure of E.coli core RNA polymerase: promoter binding and
elongation conformations of the enzyme. Cell, 83, 365–373.
Rasmussen,E.B. and Lis,J.T. (1995) Short transcripts of the ternary
complex provide insight into RNA polymerase II elongational
pausing. J. Mol. Biol., 252, 522–535.
2358
Ruet,A., Sentenac,A., Fromageot,P., Winsor,B. and Lacroute,F. (1980)
A mutation of the B220 subunit gene affects the structural and
functional properties of yeast RNA polymerase B in vitro. J. Biol.
Chem., 255, 6450–6455.
Schultz,P., Celia,H., Riva,M., Darst,S.A., Colin,P., Kornberg,R.D.,
Sentenac,A. and Oudet,P. (1990) Structural study of the yeast RNA
polymerase A: electron microscopy of lipid-bound molecules and
two-dimensional crystals. J. Mol. Biol., 216, 353–362.
Sousa,R., Chung,Y.J., Rose,J.P. and Wang,B.C. (1993) Crystal structure
of bacteriophage T7 RNA polymerase at 3.3 Å resolution. Nature,
364, 593–599.
Woychik,N.A. and Young,R.A. (1989) RNA polymerase II subunit
Rpb4 is essential for high- and low-temperature yeast cell growth.
Mol. Cell. Biol., 9, 2854–2859.
Received September 1, 1997; revised and accepted February 6, 1998